Adaptive Lookup-Table Image Processing for Low-Light Machine Vision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies, primarily designed for human visual needs, struggle to optimize image recognition in adverse conditions such as low light or bad weather, as they use fixed parameters that do not account for machine vision requirements.
Innovation Solution
An image processing method involving a weight prediction model that adjusts pixel values using lookup tables with learned mapping points, optimized for illumination and environmental conditions, and performs interpolation to enhance resolution and improve machine vision performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If fixed parameters are used for image processing to improve human-perceived quality, then ease of operation is improved, but adaptability to different environments and machine vision requirements deteriorates
Solution Approach 1:
The patent implements dynamic parameter adjustment by training neural networks to predict optimal processing parameters based on input image characteristics. The system transitions from fixed, static parameters to dynamic, adaptive parameters that automatically adjust according to environmental conditions and image content, resolving the contradiction between operational simplicity and environmental adaptability.
Solution Approach 2:
The patent changes the state of processing parameters from fixed to variable by using machine learning models to predict optimal parameter values for different imaging conditions. This allows the system to adapt parameters such as exposure compensation, gain, and processing algorithms based on detected scene characteristics, thereby improving adaptability while maintaining ease of operation through automated adjustment.
2Ease of operation
If typical image signal processing is used to improve human visual quality, then ease of operation is improved, but measurement precision for machine vision tasks deteriorates
Solution Approach 1:
The patent applies local quality by implementing different processing strategies for different regions or types of images. The system analyzes input image characteristics and applies specialized processing pipelines optimized for specific scenarios (e.g., low light, high dynamic range, portrait, landscape), thereby improving measurement precision for machine vision tasks while maintaining ease of operation through automated scene detection and pipeline selection.
Solution Approach 2:
The patent dynamically selects and adjusts processing parameters based on detected image characteristics and task requirements. By using neural networks to predict optimal parameters and dynamically switching between different processing modes, the system achieves high measurement precision for machine vision while maintaining operational simplicity through automated adaptation.
3Device complexity
If fixed parameter processing is used to simplify the system, then device complexity is reduced, but adaptability to low light and adverse conditions deteriorates
Solution Approach 1:
The patent implements self-service by enabling the image processing system to automatically analyze input characteristics and select optimal processing parameters without external intervention. The trained neural networks autonomously predict the best processing configuration for each input image, allowing the system to adapt to various conditions including low light and adverse environments while maintaining relatively simple device architecture through automated self-adjustment.
Data Source
AI summary
An image processing method including inputting an input image to a weight prediction model to predict a weight corresponding to each lookup table of a plurality of lookup tables, mapping a pixel value of the input image to a corresponding section of each lookup table of the plurality of lookup tables, calculating an adjusted pixel value corresponding to the corresponding section with respect to each lookup table of the plurality of lookup tables, and obtaining an output image with an adjusted resolution by performing a weighted sum of the adjusted pixel values according to the weight.